1. Figure 1: An example scenario. Permitted re-print
Figure 1: An example scenario. Permitted re-print.⁴¹
From: Capturing patient information at nursing shift changes: methodological evaluation of speech recognition and information extraction
J Am Med Inform Assoc. 2014;22(e1):e48-e66. doi: /amiajnl
J Am Med Inform Assoc | © The Author Published by Oxford University Press on behalf of the American Medical Informatics Association. All rights reserved. For Permissions, please

2. Figure 2: The compared microphones and recorders. 1 AUD = 0
Figure 2: The compared microphones and recorders. 1 AUD = 0.91 USD (March 23, 2014). In the soundproof professional studio, we used a professional-level recorder (EDIROL UA bit/96 kHz USB Audio Capturer) and microphone (Audio-Technica AT892cW-TH MicroSet Omnidirectional Condenser Headset). We played these recordings using professional-level speakers (EDIROL MA-15D Digital Stereo Micro Monitor Speakers) in a quiet meeting room.
From: Capturing patient information at nursing shift changes: methodological evaluation of speech recognition and information extraction
J Am Med Inform Assoc. 2014;22(e1):e48-e66. doi: /amiajnl
J Am Med Inform Assoc | © The Author Published by Oxford University Press on behalf of the American Medical Informatics Association. All rights reserved. For Permissions, please

3. Figure 3: Speech recognition correctness of the best performing device: the top bar chart details this performance on the six patient cases for the three speakers and three vocabularies over the six patient cases. The bottom bar chart addresses the differences between the six patient cases and three speakers using the best performing vocabulary (ie, nursing). With this nursing vocabulary, the average percentages of correctly recognized, substituted, deleted, and inserted words across the six patient cases were 70.5 (64.4) [61.9], 20.2 (22.0) [28.5], 9.3 (13.6) [9.5], and 2.7 (3.1) [1.8] for MN (FN) [FA]. MN, native male; FN, native female; FA, accented female.
From: Capturing patient information at nursing shift changes: methodological evaluation of speech recognition and information extraction
J Am Med Inform Assoc. 2014;22(e1):e48-e66. doi: /amiajnl
J Am Med Inform Assoc | © The Author Published by Oxford University Press on behalf of the American Medical Informatics Association. All rights reserved. For Permissions, please

4. Figure 4: Precision (4a), recall (4b), and F1 (4c) percentages and learning curves for different cross-validation (CV) settings (ie, training set sizes of 30, 60, 90, 120, and 149 (ie, leave-one-out [LOO] CV) reports with mutually exclusive folds that in combination cover all data). NA refers to the category for irrelevant text. The horizontal direction of the histograms reflects the contribution of having more data for training and the vertical direction the effects of the coupled measured of precision and recall in F1 (see the glossary in online supplementary appendix).
From: Capturing patient information at nursing shift changes: methodological evaluation of speech recognition and information extraction
J Am Med Inform Assoc. 2014;22(e1):e48-e66. doi: /amiajnl
J Am Med Inform Assoc | © The Author Published by Oxford University Press on behalf of the American Medical Informatics Association. All rights reserved. For Permissions, please